search

lshqqytiger / ZLUDA

CUDA on AMD GPUs
Apache License 2.0
250 stars 32 forks source link

ollama with zluda on 7780hs 780m AMD #13

Open kenhuang opened 4 months ago

kenhuang commented 4 months ago

confirm working using zluda + stable diffusion via GPU here

C:\Users\ken>d:\ZLUDA\zluda.exe -- ollama.exe serve
time=2024-04-22T20:03:03.214+10:00 level=INFO source=images.go:817 msg="total blobs: 20"
time=2024-04-22T20:03:03.216+10:00 level=INFO source=images.go:824 msg="total unused blobs removed: 0"
time=2024-04-22T20:03:03.217+10:00 level=INFO source=routes.go:1143 msg="Listening on [::]:11434 (version 0.1.32)"
time=2024-04-22T20:03:03.220+10:00 level=INFO source=payload.go:28 msg="extracting embedded files" dir=C:\Users\ken\AppData\Local\Temp\ollama3073857994\runners
time=2024-04-22T20:03:03.428+10:00 level=INFO source=payload.go:41 msg="Dynamic LLM libraries [cpu cpu_avx cpu_avx2 cuda_v11.3 rocm_v5.7]"
[GIN] 2024/04/22 - 20:03:08 | 200 |            0s |       127.0.0.1 | HEAD     "/"
[GIN] 2024/04/22 - 20:03:08 | 200 |       535.6µs |       127.0.0.1 | POST     "/api/show"
time=2024-04-22T20:03:08.550+10:00 level=INFO source=gpu.go:121 msg="Detecting GPU type"
time=2024-04-22T20:03:08.551+10:00 level=INFO source=gpu.go:268 msg="Searching for GPU management library cudart64_*.dll"
time=2024-04-22T20:03:08.583+10:00 level=INFO source=gpu.go:314 msg="Discovered GPU libraries: [C:\\Users\\ken\\AppData\\Local\\Programs\\Ollama\\cudart64_110.dll C:\\Program Files (x86)\\NVIDIA Corporation\\PhysX\\Common\\cudart64_60.dll]"
time=2024-04-22T20:03:08.622+10:00 level=INFO source=gpu.go:126 msg="Nvidia GPU detected via cudart"
time=2024-04-22T20:03:08.622+10:00 level=INFO source=cpu_common.go:11 msg="CPU has AVX2"
time=2024-04-22T20:03:08.846+10:00 level=INFO source=gpu.go:202 msg="[cudart] CUDART CUDA Compute Capability detected: 8.8"
time=2024-04-22T20:03:08.846+10:00 level=INFO source=gpu.go:121 msg="Detecting GPU type"
time=2024-04-22T20:03:08.846+10:00 level=INFO source=gpu.go:268 msg="Searching for GPU management library cudart64_*.dll"
time=2024-04-22T20:03:08.878+10:00 level=INFO source=gpu.go:314 msg="Discovered GPU libraries: [C:\\Users\\ken\\AppData\\Local\\Programs\\Ollama\\cudart64_110.dll C:\\Program Files (x86)\\NVIDIA Corporation\\PhysX\\Common\\cudart64_60.dll]"
time=2024-04-22T20:03:08.879+10:00 level=INFO source=gpu.go:126 msg="Nvidia GPU detected via cudart"
time=2024-04-22T20:03:08.879+10:00 level=INFO source=cpu_common.go:11 msg="CPU has AVX2"
time=2024-04-22T20:03:08.879+10:00 level=INFO source=gpu.go:202 msg="[cudart] CUDART CUDA Compute Capability detected: 8.8"
time=2024-04-22T20:03:08.879+10:00 level=INFO source=server.go:127 msg="offload to gpu" reallayers=33 layers=33 required="5222.5 MiB" used="5222.5 MiB" available="22085.4 MiB" kv="1024.0 MiB" fulloffload="164.0 MiB" partialoffload="193.0 MiB"
time=2024-04-22T20:03:08.879+10:00 level=INFO source=cpu_common.go:11 msg="CPU has AVX2"
time=2024-04-22T20:03:08.885+10:00 level=INFO source=server.go:264 msg="starting llama server" cmd="C:\\Users\\ken\\AppData\\Local\\Temp\\ollama3073857994\\runners\\cuda_v11.3\\ollama_llama_server.exe --model C:\\Users\\ken\\.ollama\\models\\blobs\\sha256-8934d96d3f08982e95922b2b7a2c626a1fe873d7c3b06e8e56d7bc0a1fef9246 --ctx-size 2048 --batch-size 512 --embedding --log-disable --n-gpu-layers 33 --port 63294"
time=2024-04-22T20:03:08.911+10:00 level=INFO source=server.go:389 msg="waiting for llama runner to start responding"
{"function":"server_params_parse","level":"INFO","line":2603,"msg":"logging to file is disabled.","tid":"17584","timestamp":1713780189}
{"build":2679,"commit":"7593639","function":"wmain","level":"INFO","line":2820,"msg":"build info","tid":"17584","timestamp":1713780189}
{"function":"wmain","level":"INFO","line":2827,"msg":"system info","n_threads":8,"n_threads_batch":-1,"system_info":"AVX = 1 | AVX_VNNI = 0 | AVX2 = 0 | AVX512 = 0 | AVX512_VBMI = 0 | AVX512_VNNI = 0 | FMA = 0 | NEON = 0 | ARM_FMA = 0 | F16C = 0 | FP16_VA = 0 | WASM_SIMD = 0 | BLAS = 1 | SSE3 = 0 | SSSE3 = 0 | VSX = 0 | MATMUL_INT8 = 0 | ","tid":"17584","timestamp":1713780189,"total_threads":16}
llama_model_loader: loaded meta data with 23 key-value pairs and 291 tensors from C:\Users\ken\.ollama\models\blobs\sha256-8934d96d3f08982e95922b2b7a2c626a1fe873d7c3b06e8e56d7bc0a1fef9246 (version GGUF V3 (latest))
llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
llama_model_loader: - kv   0:                       general.architecture str              = llama
llama_model_loader: - kv   1:                               general.name str              = LLaMA v2
llama_model_loader: - kv   2:                       llama.context_length u32              = 4096
llama_model_loader: - kv   3:                     llama.embedding_length u32              = 4096
llama_model_loader: - kv   4:                          llama.block_count u32              = 32
llama_model_loader: - kv   5:                  llama.feed_forward_length u32              = 11008
llama_model_loader: - kv   6:                 llama.rope.dimension_count u32              = 128
llama_model_loader: - kv   7:                 llama.attention.head_count u32              = 32
llama_model_loader: - kv   8:              llama.attention.head_count_kv u32              = 32
llama_model_loader: - kv   9:     llama.attention.layer_norm_rms_epsilon f32              = 0.000010
llama_model_loader: - kv  10:                          general.file_type u32              = 2
llama_model_loader: - kv  11:                       tokenizer.ggml.model str              = llama
llama_model_loader: - kv  12:                      tokenizer.ggml.tokens arr[str,32000]   = ["<unk>", "<s>", "</s>", "<0x00>", "<...
llama_model_loader: - kv  13:                      tokenizer.ggml.scores arr[f32,32000]   = [0.000000, 0.000000, 0.000000, 0.0000...
llama_model_loader: - kv  14:                  tokenizer.ggml.token_type arr[i32,32000]   = [2, 3, 3, 6, 6, 6, 6, 6, 6, 6, 6, 6, ...
llama_model_loader: - kv  15:                      tokenizer.ggml.merges arr[str,61249]   = ["▁ t", "e r", "i n", "▁ a", "e n...
llama_model_loader: - kv  16:                tokenizer.ggml.bos_token_id u32              = 1
llama_model_loader: - kv  17:                tokenizer.ggml.eos_token_id u32              = 2
llama_model_loader: - kv  18:            tokenizer.ggml.unknown_token_id u32              = 0
llama_model_loader: - kv  19:               tokenizer.ggml.add_bos_token bool             = true
llama_model_loader: - kv  20:               tokenizer.ggml.add_eos_token bool             = false
llama_model_loader: - kv  21:                    tokenizer.chat_template str              = {% if messages[0]['role'] == 'system'...
llama_model_loader: - kv  22:               general.quantization_version u32              = 2
llama_model_loader: - type  f32:   65 tensors
llama_model_loader: - type q4_0:  225 tensors
llama_model_loader: - type q6_K:    1 tensors
llm_load_vocab: special tokens definition check successful ( 259/32000 ).
llm_load_print_meta: format           = GGUF V3 (latest)
llm_load_print_meta: arch             = llama
llm_load_print_meta: vocab type       = SPM
llm_load_print_meta: n_vocab          = 32000
llm_load_print_meta: n_merges         = 0
llm_load_print_meta: n_ctx_train      = 4096
llm_load_print_meta: n_embd           = 4096
llm_load_print_meta: n_head           = 32
llm_load_print_meta: n_head_kv        = 32
llm_load_print_meta: n_layer          = 32
llm_load_print_meta: n_rot            = 128
llm_load_print_meta: n_embd_head_k    = 128
llm_load_print_meta: n_embd_head_v    = 128
llm_load_print_meta: n_gqa            = 1
llm_load_print_meta: n_embd_k_gqa     = 4096
llm_load_print_meta: n_embd_v_gqa     = 4096
llm_load_print_meta: f_norm_eps       = 0.0e+00
llm_load_print_meta: f_norm_rms_eps   = 1.0e-05
llm_load_print_meta: f_clamp_kqv      = 0.0e+00
llm_load_print_meta: f_max_alibi_bias = 0.0e+00
llm_load_print_meta: f_logit_scale    = 0.0e+00
llm_load_print_meta: n_ff             = 11008
llm_load_print_meta: n_expert         = 0
llm_load_print_meta: n_expert_used    = 0
llm_load_print_meta: causal attn      = 1
llm_load_print_meta: pooling type     = 0
llm_load_print_meta: rope type        = 0
llm_load_print_meta: rope scaling     = linear
llm_load_print_meta: freq_base_train  = 10000.0
llm_load_print_meta: freq_scale_train = 1
llm_load_print_meta: n_yarn_orig_ctx  = 4096
llm_load_print_meta: rope_finetuned   = unknown
llm_load_print_meta: ssm_d_conv       = 0
llm_load_print_meta: ssm_d_inner      = 0
llm_load_print_meta: ssm_d_state      = 0
llm_load_print_meta: ssm_dt_rank      = 0
llm_load_print_meta: model type       = 7B
llm_load_print_meta: model ftype      = Q4_0
llm_load_print_meta: model params     = 6.74 B
llm_load_print_meta: model size       = 3.56 GiB (4.54 BPW)
llm_load_print_meta: general.name     = LLaMA v2
llm_load_print_meta: BOS token        = 1 '<s>'
llm_load_print_meta: EOS token        = 2 '</s>'
llm_load_print_meta: UNK token        = 0 '<unk>'
llm_load_print_meta: LF token         = 13 '<0x0A>'
ggml_cuda_init: GGML_CUDA_FORCE_MMQ:   no
ggml_cuda_init: CUDA_USE_TENSOR_CORES: yes
ggml_cuda_init: found 1 CUDA devices:
  Device 0: AMD Radeon 780M Graphics [ZLUDA], compute capability 8.8, VMM: no
llm_load_tensors: ggml ctx size =    0.22 MiB
llm_load_tensors: offloading 32 repeating layers to GPU
llm_load_tensors: offloading non-repeating layers to GPU
llm_load_tensors: offloaded 33/33 layers to GPU
llm_load_tensors:        CPU buffer size =    70.31 MiB
llm_load_tensors:      CUDA0 buffer size =  3577.56 MiB
..................................................................................................
llama_new_context_with_model: n_ctx      = 2048
llama_new_context_with_model: n_batch    = 512
llama_new_context_with_model: n_ubatch   = 512
llama_new_context_with_model: freq_base  = 10000.0
llama_new_context_with_model: freq_scale = 1
llama_kv_cache_init:      CUDA0 KV buffer size =  1024.00 MiB
llama_new_context_with_model: KV self size  = 1024.00 MiB, K (f16):  512.00 MiB, V (f16):  512.00 MiB
llama_new_context_with_model:  CUDA_Host  output buffer size =     0.14 MiB
llama_new_context_with_model:      CUDA0 compute buffer size =   164.00 MiB
llama_new_context_with_model:  CUDA_Host compute buffer size =    12.01 MiB
llama_new_context_with_model: graph nodes  = 1030
llama_new_context_with_model: graph splits = 2
CUDA error: CUBLAS_STATUS_NOT_INITIALIZED
  current device: 0, in function cublas_handle at C:\a\ollama\ollama\llm\llama.cpp\ggml-cuda/common.cuh:526
  cublasCreate_v2(&cublas_handles[device])
GGML_ASSERT: C:\a\ollama\ollama\llm\llama.cpp\ggml-cuda.cu:60: !"CUDA error"
time=2024-04-22T20:03:13.722+10:00 level=ERROR source=routes.go:120 msg="error loading llama server" error="llama runner process no longer running: 3221226505 CUDA error\""
[GIN] 2024/04/22 - 20:03:13 | 500 |     5.562023s |       127.0.0.1 | POST     "/api/generate"
^C
C:\Users\ken>d:\ZLUDA\zluda.exe -- ollama.exe serve
time=2024-04-22T20:11:47.288+10:00 level=INFO source=images.go:817 msg="total blobs: 20"
time=2024-04-22T20:11:47.289+10:00 level=INFO source=images.go:824 msg="total unused blobs removed: 0"
time=2024-04-22T20:11:47.290+10:00 level=INFO source=routes.go:1143 msg="Listening on [::]:11434 (version 0.1.32)"
time=2024-04-22T20:11:47.292+10:00 level=INFO source=payload.go:28 msg="extracting embedded files" dir=C:\Users\ken\AppData\Local\Temp\ollama1878556081\runners
time=2024-04-22T20:11:47.483+10:00 level=INFO source=payload.go:41 msg="Dynamic LLM libraries [cpu_avx2 cuda_v11.3 rocm_v5.7 cpu cpu_avx]"
[GIN] 2024/04/22 - 20:11:52 | 200 |            0s |       127.0.0.1 | HEAD     "/"
[GIN] 2024/04/22 - 20:11:52 | 200 |       530.1µs |       127.0.0.1 | POST     "/api/show"
time=2024-04-22T20:11:52.991+10:00 level=INFO source=gpu.go:121 msg="Detecting GPU type"
time=2024-04-22T20:11:52.991+10:00 level=INFO source=gpu.go:268 msg="Searching for GPU management library cudart64_*.dll"
time=2024-04-22T20:11:53.032+10:00 level=INFO source=gpu.go:314 msg="Discovered GPU libraries: [C:\\Users\\ken\\AppData\\Local\\Programs\\Ollama\\cudart64_110.dll C:\\Program Files (x86)\\NVIDIA Corporation\\PhysX\\Common\\cudart64_60.dll]"
time=2024-04-22T20:11:53.058+10:00 level=INFO source=gpu.go:126 msg="Nvidia GPU detected via cudart"
time=2024-04-22T20:11:53.058+10:00 level=INFO source=cpu_common.go:11 msg="CPU has AVX2"
time=2024-04-22T20:11:53.281+10:00 level=INFO source=gpu.go:202 msg="[cudart] CUDART CUDA Compute Capability detected: 8.8"
time=2024-04-22T20:11:53.281+10:00 level=INFO source=gpu.go:121 msg="Detecting GPU type"
time=2024-04-22T20:11:53.281+10:00 level=INFO source=gpu.go:268 msg="Searching for GPU management library cudart64_*.dll"
time=2024-04-22T20:11:53.313+10:00 level=INFO source=gpu.go:314 msg="Discovered GPU libraries: [C:\\Users\\ken\\AppData\\Local\\Programs\\Ollama\\cudart64_110.dll C:\\Program Files (x86)\\NVIDIA Corporation\\PhysX\\Common\\cudart64_60.dll]"
time=2024-04-22T20:11:53.314+10:00 level=INFO source=gpu.go:126 msg="Nvidia GPU detected via cudart"
time=2024-04-22T20:11:53.314+10:00 level=INFO source=cpu_common.go:11 msg="CPU has AVX2"
time=2024-04-22T20:11:53.315+10:00 level=INFO source=gpu.go:202 msg="[cudart] CUDART CUDA Compute Capability detected: 8.8"
time=2024-04-22T20:11:53.315+10:00 level=INFO source=server.go:127 msg="offload to gpu" reallayers=33 layers=33 required="5222.5 MiB" used="5222.5 MiB" available="22085.4 MiB" kv="1024.0 MiB" fulloffload="164.0 MiB" partialoffload="193.0 MiB"
time=2024-04-22T20:11:53.315+10:00 level=INFO source=cpu_common.go:11 msg="CPU has AVX2"
time=2024-04-22T20:11:53.320+10:00 level=INFO source=server.go:264 msg="starting llama server" cmd="C:\\Users\\ken\\AppData\\Local\\Temp\\ollama1878556081\\runners\\cuda_v11.3\\ollama_llama_server.exe --model C:\\Users\\ken\\.ollama\\models\\blobs\\sha256-8934d96d3f08982e95922b2b7a2c626a1fe873d7c3b06e8e56d7bc0a1fef9246 --ctx-size 2048 --batch-size 512 --embedding --log-disable --n-gpu-layers 33 --port 63561"
time=2024-04-22T20:11:53.345+10:00 level=INFO source=server.go:389 msg="waiting for llama runner to start responding"
{"function":"server_params_parse","level":"INFO","line":2603,"msg":"logging to file is disabled.","tid":"1600","timestamp":1713780713}
{"build":2679,"commit":"7593639","function":"wmain","level":"INFO","line":2820,"msg":"build info","tid":"1600","timestamp":1713780713}
{"function":"wmain","level":"INFO","line":2827,"msg":"system info","n_threads":8,"n_threads_batch":-1,"system_info":"AVX = 1 | AVX_VNNI = 0 | AVX2 = 0 | AVX512 = 0 | AVX512_VBMI = 0 | AVX512_VNNI = 0 | FMA = 0 | NEON = 0 | ARM_FMA = 0 | F16C = 0 | FP16_VA = 0 | WASM_SIMD = 0 | BLAS = 1 | SSE3 = 0 | SSSE3 = 0 | VSX = 0 | MATMUL_INT8 = 0 | ","tid":"1600","timestamp":1713780713,"total_threads":16}
llama_model_loader: loaded meta data with 23 key-value pairs and 291 tensors from C:\Users\ken\.ollama\models\blobs\sha256-8934d96d3f08982e95922b2b7a2c626a1fe873d7c3b06e8e56d7bc0a1fef9246 (version GGUF V3 (latest))
llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
llama_model_loader: - kv   0:                       general.architecture str              = llama
llama_model_loader: - kv   1:                               general.name str              = LLaMA v2
llama_model_loader: - kv   2:                       llama.context_length u32              = 4096
llama_model_loader: - kv   3:                     llama.embedding_length u32              = 4096
llama_model_loader: - kv   4:                          llama.block_count u32              = 32
llama_model_loader: - kv   5:                  llama.feed_forward_length u32              = 11008
llama_model_loader: - kv   6:                 llama.rope.dimension_count u32              = 128
llama_model_loader: - kv   7:                 llama.attention.head_count u32              = 32
llama_model_loader: - kv   8:              llama.attention.head_count_kv u32              = 32
llama_model_loader: - kv   9:     llama.attention.layer_norm_rms_epsilon f32              = 0.000010
llama_model_loader: - kv  10:                          general.file_type u32              = 2
llama_model_loader: - kv  11:                       tokenizer.ggml.model str              = llama
llama_model_loader: - kv  12:                      tokenizer.ggml.tokens arr[str,32000]   = ["<unk>", "<s>", "</s>", "<0x00>", "<...
llama_model_loader: - kv  13:                      tokenizer.ggml.scores arr[f32,32000]   = [0.000000, 0.000000, 0.000000, 0.0000...
llama_model_loader: - kv  14:                  tokenizer.ggml.token_type arr[i32,32000]   = [2, 3, 3, 6, 6, 6, 6, 6, 6, 6, 6, 6, ...
llama_model_loader: - kv  15:                      tokenizer.ggml.merges arr[str,61249]   = ["▁ t", "e r", "i n", "▁ a", "e n...
llama_model_loader: - kv  16:                tokenizer.ggml.bos_token_id u32              = 1
llama_model_loader: - kv  17:                tokenizer.ggml.eos_token_id u32              = 2
llama_model_loader: - kv  18:            tokenizer.ggml.unknown_token_id u32              = 0
llama_model_loader: - kv  19:               tokenizer.ggml.add_bos_token bool             = true
llama_model_loader: - kv  20:               tokenizer.ggml.add_eos_token bool             = false
llama_model_loader: - kv  21:                    tokenizer.chat_template str              = {% if messages[0]['role'] == 'system'...
llama_model_loader: - kv  22:               general.quantization_version u32              = 2
llama_model_loader: - type  f32:   65 tensors
llama_model_loader: - type q4_0:  225 tensors
llama_model_loader: - type q6_K:    1 tensors
llm_load_vocab: special tokens definition check successful ( 259/32000 ).
llm_load_print_meta: format           = GGUF V3 (latest)
llm_load_print_meta: arch             = llama
llm_load_print_meta: vocab type       = SPM
llm_load_print_meta: n_vocab          = 32000
llm_load_print_meta: n_merges         = 0
llm_load_print_meta: n_ctx_train      = 4096
llm_load_print_meta: n_embd           = 4096
llm_load_print_meta: n_head           = 32
llm_load_print_meta: n_head_kv        = 32
llm_load_print_meta: n_layer          = 32
llm_load_print_meta: n_rot            = 128
llm_load_print_meta: n_embd_head_k    = 128
llm_load_print_meta: n_embd_head_v    = 128
llm_load_print_meta: n_gqa            = 1
llm_load_print_meta: n_embd_k_gqa     = 4096
llm_load_print_meta: n_embd_v_gqa     = 4096
llm_load_print_meta: f_norm_eps       = 0.0e+00
llm_load_print_meta: f_norm_rms_eps   = 1.0e-05
llm_load_print_meta: f_clamp_kqv      = 0.0e+00
llm_load_print_meta: f_max_alibi_bias = 0.0e+00
llm_load_print_meta: f_logit_scale    = 0.0e+00
llm_load_print_meta: n_ff             = 11008
llm_load_print_meta: n_expert         = 0
llm_load_print_meta: n_expert_used    = 0
llm_load_print_meta: causal attn      = 1
llm_load_print_meta: pooling type     = 0
llm_load_print_meta: rope type        = 0
llm_load_print_meta: rope scaling     = linear
llm_load_print_meta: freq_base_train  = 10000.0
llm_load_print_meta: freq_scale_train = 1
llm_load_print_meta: n_yarn_orig_ctx  = 4096
llm_load_print_meta: rope_finetuned   = unknown
llm_load_print_meta: ssm_d_conv       = 0
llm_load_print_meta: ssm_d_inner      = 0
llm_load_print_meta: ssm_d_state      = 0
llm_load_print_meta: ssm_dt_rank      = 0
llm_load_print_meta: model type       = 7B
llm_load_print_meta: model ftype      = Q4_0
llm_load_print_meta: model params     = 6.74 B
llm_load_print_meta: model size       = 3.56 GiB (4.54 BPW)
llm_load_print_meta: general.name     = LLaMA v2
llm_load_print_meta: BOS token        = 1 '<s>'
llm_load_print_meta: EOS token        = 2 '</s>'
llm_load_print_meta: UNK token        = 0 '<unk>'
llm_load_print_meta: LF token         = 13 '<0x0A>'
ggml_cuda_init: GGML_CUDA_FORCE_MMQ:   no
ggml_cuda_init: CUDA_USE_TENSOR_CORES: yes
ggml_cuda_init: found 1 CUDA devices:
  Device 0: AMD Radeon 780M Graphics [ZLUDA], compute capability 8.8, VMM: no
llm_load_tensors: ggml ctx size =    0.22 MiB
llm_load_tensors: offloading 32 repeating layers to GPU
llm_load_tensors: offloading non-repeating layers to GPU
llm_load_tensors: offloaded 33/33 layers to GPU
llm_load_tensors:        CPU buffer size =    70.31 MiB
llm_load_tensors:      CUDA0 buffer size =  3577.56 MiB
..................................................................................................
llama_new_context_with_model: n_ctx      = 2048
llama_new_context_with_model: n_batch    = 512
llama_new_context_with_model: n_ubatch   = 512
llama_new_context_with_model: freq_base  = 10000.0
llama_new_context_with_model: freq_scale = 1
llama_kv_cache_init:      CUDA0 KV buffer size =  1024.00 MiB
llama_new_context_with_model: KV self size  = 1024.00 MiB, K (f16):  512.00 MiB, V (f16):  512.00 MiB
llama_new_context_with_model:  CUDA_Host  output buffer size =     0.14 MiB
llama_new_context_with_model:      CUDA0 compute buffer size =   164.00 MiB
llama_new_context_with_model:  CUDA_Host compute buffer size =    12.01 MiB
llama_new_context_with_model: graph nodes  = 1030
llama_new_context_with_model: graph splits = 2
CUDA error: CUBLAS_STATUS_NOT_INITIALIZED
  current device: 0, in function cublas_handle at C:\a\ollama\ollama\llm\llama.cpp\ggml-cuda/common.cuh:526
  cublasCreate_v2(&cublas_handles[device])
GGML_ASSERT: C:\a\ollama\ollama\llm\llama.cpp\ggml-cuda.cu:60: !"CUDA error"
time=2024-04-22T20:12:00.555+10:00 level=ERROR source=routes.go:120 msg="error loading llama server" error="llama runner process no longer running: 3221226505 CUDA error\""
[GIN] 2024/04/22 - 20:12:00 | 500 |    7.9525007s |       127.0.0.1 | POST     "/api/generate"
kenhuang commented 4 months ago

error without zluda

C:\Users\ken>ollama.exe serve
time=2024-04-22T20:19:08.641+10:00 level=INFO source=images.go:817 msg="total blobs: 20"
time=2024-04-22T20:19:08.642+10:00 level=INFO source=images.go:824 msg="total unused blobs removed: 0"
time=2024-04-22T20:19:08.643+10:00 level=INFO source=routes.go:1143 msg="Listening on [::]:11434 (version 0.1.32)"
time=2024-04-22T20:19:08.654+10:00 level=INFO source=payload.go:28 msg="extracting embedded files" dir=C:\Users\ken\AppData\Local\Temp\ollama570398800\runners
time=2024-04-22T20:19:08.850+10:00 level=INFO source=payload.go:41 msg="Dynamic LLM libraries [cpu_avx2 cuda_v11.3 rocm_v5.7 cpu cpu_avx]"
[GIN] 2024/04/22 - 20:19:14 | 200 |            0s |       127.0.0.1 | HEAD     "/"
[GIN] 2024/04/22 - 20:19:14 | 200 |      1.0618ms |       127.0.0.1 | POST     "/api/show"
time=2024-04-22T20:19:14.572+10:00 level=INFO source=gpu.go:121 msg="Detecting GPU type"
time=2024-04-22T20:19:14.572+10:00 level=INFO source=gpu.go:268 msg="Searching for GPU management library cudart64_*.dll"
time=2024-04-22T20:19:14.611+10:00 level=INFO source=gpu.go:314 msg="Discovered GPU libraries: [C:\\Users\\ken\\AppData\\Local\\Programs\\Ollama\\cudart64_110.dll C:\\Program Files (x86)\\NVIDIA Corporation\\PhysX\\Common\\cudart64_60.dll]"
time=2024-04-22T20:19:14.612+10:00 level=INFO source=gpu.go:343 msg="Unable to load cudart CUDA management library C:\\Users\\ken\\AppData\\Local\\Programs\\Ollama\\cudart64_110.dll: your nvidia driver is too old or missing, please upgrade to run ollama"
time=2024-04-22T20:19:14.612+10:00 level=INFO source=gpu.go:343 msg="Unable to load cudart CUDA management library C:\\Program Files (x86)\\NVIDIA Corporation\\PhysX\\Common\\cudart64_60.dll: your nvidia driver is too old or missing, please upgrade to run ollama"
time=2024-04-22T20:19:14.612+10:00 level=INFO source=gpu.go:268 msg="Searching for GPU management library nvml.dll"
time=2024-04-22T20:19:14.646+10:00 level=INFO source=gpu.go:314 msg="Discovered GPU libraries: []"
time=2024-04-22T20:19:14.646+10:00 level=INFO source=cpu_common.go:11 msg="CPU has AVX2"
time=2024-04-22T20:19:14.663+10:00 level=INFO source=amd_windows.go:40 msg="AMD Driver: 50732000"
time=2024-04-22T20:19:14.663+10:00 level=INFO source=amd_windows.go:69 msg="detected 1 hip devices"
time=2024-04-22T20:19:14.663+10:00 level=INFO source=amd_windows.go:87 msg="[0] Name: AMD Radeon 780M Graphics"
time=2024-04-22T20:19:14.663+10:00 level=INFO source=amd_windows.go:90 msg="[0] GcnArchName: gfx1103"
time=2024-04-22T20:19:14.900+10:00 level=INFO source=amd_windows.go:117 msg="[0] Total Mem: 23158226944"
time=2024-04-22T20:19:14.900+10:00 level=INFO source=amd_windows.go:118 msg="[0] Free Mem:  23309320192"
time=2024-04-22T20:19:14.900+10:00 level=INFO source=assets.go:123 msg="Updating PATH to C:\\Users\\ken\\AppData\\Local\\Programs\\Ollama\\rocm;C:\\Program Files (x86)\\NVIDIA Corporation\\PhysX\\Common;C:\\WINDOWS\\system32;C:\\WINDOWS;C:\\WINDOWS\\System32\\Wbem;C:\\WINDOWS\\System32\\WindowsPowerShell\\v1.0\\;C:\\WINDOWS\\System32\\OpenSSH\\;C:\\Recovery\\OEM\\Backup\\;C:\\Program Files\\dotnet\\;C:\\Program Files\\Git\\cmd;C:\\Program Files\\Git LFS;C:\\WINDOWS\\system32;C:\\WINDOWS;C:\\WINDOWS\\System32\\Wbem;C:\\WINDOWS\\System32\\WindowsPowerShell\\v1.0\\;C:\\WINDOWS\\System32\\OpenSSH\\;C:\\Program Files\\Docker\\Docker\\resources\\bin;C:\\Program Files\\nodejs\\;C:\\Program Files (x86)\\NVIDIA Corporation\\PhysX\\Common;C:\\WINDOWS\\system32;C:\\WINDOWS;C:\\WINDOWS\\System32\\Wbem;C:\\WINDOWS\\System32\\WindowsPowerShell\\v1.0\\;C:\\WINDOWS\\System32\\OpenSSH\\;C:\\Recovery\\OEM\\Backup\\;C:\\Program Files\\dotnet\\;C:\\Program Files\\Git\\cmd;C:\\Program Files\\Git LFS;C:\\WINDOWS\\system32;C:\\WINDOWS;C:\\WINDOWS\\System32\\Wbem;C:\\WINDOWS\\System32\\WindowsPowerShell\\v1.0\\;C:\\WINDOWS\\System32\\OpenSSH\\;C:\\Program Files\\Docker\\Docker\\resources\\bin;C:\\Program Files\\nodejs\\;C:\\Users\\ken\\AppData\\Local\\Programs\\Python\\Python310\\Scripts\\;C:\\Users\\ken\\AppData\\Local\\Programs\\Python\\Python310\\;C:\\Users\\ken\\AppData\\Local\\Microsoft\\WindowsApps;C:\\Users\\ken\\AppData\\Local\\Programs\\Microsoft VS Code\\bin;C:\\Users\\ken\\AppData\\Roaming\\npm;C:\\Program Files\\AMD\\ROCm\\5.7\\bin;;"
time=2024-04-22T20:19:14.962+10:00 level=INFO source=gpu.go:121 msg="Detecting GPU type"
time=2024-04-22T20:19:14.962+10:00 level=INFO source=gpu.go:268 msg="Searching for GPU management library cudart64_*.dll"
time=2024-04-22T20:19:14.997+10:00 level=INFO source=gpu.go:314 msg="Discovered GPU libraries: [C:\\Users\\ken\\AppData\\Local\\Programs\\Ollama\\cudart64_110.dll C:\\Program Files (x86)\\NVIDIA Corporation\\PhysX\\Common\\cudart64_60.dll]"
time=2024-04-22T20:19:14.998+10:00 level=INFO source=gpu.go:343 msg="Unable to load cudart CUDA management library C:\\Users\\ken\\AppData\\Local\\Programs\\Ollama\\cudart64_110.dll: your nvidia driver is too old or missing, please upgrade to run ollama"
time=2024-04-22T20:19:14.998+10:00 level=INFO source=gpu.go:343 msg="Unable to load cudart CUDA management library C:\\Program Files (x86)\\NVIDIA Corporation\\PhysX\\Common\\cudart64_60.dll: your nvidia driver is too old or missing, please upgrade to run ollama"
time=2024-04-22T20:19:14.999+10:00 level=INFO source=gpu.go:268 msg="Searching for GPU management library nvml.dll"
time=2024-04-22T20:19:15.035+10:00 level=INFO source=gpu.go:314 msg="Discovered GPU libraries: []"
time=2024-04-22T20:19:15.035+10:00 level=INFO source=cpu_common.go:11 msg="CPU has AVX2"
time=2024-04-22T20:19:15.046+10:00 level=INFO source=amd_windows.go:40 msg="AMD Driver: 50732000"
time=2024-04-22T20:19:15.046+10:00 level=INFO source=amd_windows.go:69 msg="detected 1 hip devices"
time=2024-04-22T20:19:15.046+10:00 level=INFO source=amd_windows.go:87 msg="[0] Name: AMD Radeon 780M Graphics"
time=2024-04-22T20:19:15.047+10:00 level=INFO source=amd_windows.go:90 msg="[0] GcnArchName: gfx1103"
time=2024-04-22T20:19:15.268+10:00 level=INFO source=amd_windows.go:117 msg="[0] Total Mem: 23158226944"
time=2024-04-22T20:19:15.268+10:00 level=INFO source=amd_windows.go:118 msg="[0] Free Mem:  23309320192"
time=2024-04-22T20:19:15.323+10:00 level=INFO source=server.go:127 msg="offload to gpu" reallayers=33 layers=33 required="5222.5 MiB" used="5222.5 MiB" available="22229.5 MiB" kv="1024.0 MiB" fulloffload="164.0 MiB" partialoffload="193.0 MiB"
time=2024-04-22T20:19:15.323+10:00 level=INFO source=cpu_common.go:11 msg="CPU has AVX2"
time=2024-04-22T20:19:15.330+10:00 level=INFO source=server.go:264 msg="starting llama server" cmd="C:\\Users\\ken\\AppData\\Local\\Temp\\ollama570398800\\runners\\rocm_v5.7\\ollama_llama_server.exe --model C:\\Users\\ken\\.ollama\\models\\blobs\\sha256-8934d96d3f08982e95922b2b7a2c626a1fe873d7c3b06e8e56d7bc0a1fef9246 --ctx-size 2048 --batch-size 512 --embedding --log-disable --n-gpu-layers 33 --port 63771"
time=2024-04-22T20:19:15.350+10:00 level=INFO source=server.go:389 msg="waiting for llama runner to start responding"
{"function":"server_params_parse","level":"INFO","line":2603,"msg":"logging to file is disabled.","tid":"18916","timestamp":1713781155}
{"build":2679,"commit":"7593639","function":"wmain","level":"INFO","line":2820,"msg":"build info","tid":"18916","timestamp":1713781155}
{"function":"wmain","level":"INFO","line":2827,"msg":"system info","n_threads":8,"n_threads_batch":-1,"system_info":"AVX = 1 | AVX_VNNI = 0 | AVX2 = 0 | AVX512 = 0 | AVX512_VBMI = 0 | AVX512_VNNI = 0 | FMA = 1 | NEON = 0 | ARM_FMA = 0 | F16C = 1 | FP16_VA = 0 | WASM_SIMD = 0 | BLAS = 1 | SSE3 = 1 | SSSE3 = 1 | VSX = 0 | MATMUL_INT8 = 0 | ","tid":"18916","timestamp":1713781155,"total_threads":16}
llama_model_loader: loaded meta data with 23 key-value pairs and 291 tensors from C:\Users\ken\.ollama\models\blobs\sha256-8934d96d3f08982e95922b2b7a2c626a1fe873d7c3b06e8e56d7bc0a1fef9246 (version GGUF V3 (latest))
llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
llama_model_loader: - kv   0:                       general.architecture str              = llama
llama_model_loader: - kv   1:                               general.name str              = LLaMA v2
llama_model_loader: - kv   2:                       llama.context_length u32              = 4096
llama_model_loader: - kv   3:                     llama.embedding_length u32              = 4096
llama_model_loader: - kv   4:                          llama.block_count u32              = 32
llama_model_loader: - kv   5:                  llama.feed_forward_length u32              = 11008
llama_model_loader: - kv   6:                 llama.rope.dimension_count u32              = 128
llama_model_loader: - kv   7:                 llama.attention.head_count u32              = 32
llama_model_loader: - kv   8:              llama.attention.head_count_kv u32              = 32
llama_model_loader: - kv   9:     llama.attention.layer_norm_rms_epsilon f32              = 0.000010
llama_model_loader: - kv  10:                          general.file_type u32              = 2
llama_model_loader: - kv  11:                       tokenizer.ggml.model str              = llama
llama_model_loader: - kv  12:                      tokenizer.ggml.tokens arr[str,32000]   = ["<unk>", "<s>", "</s>", "<0x00>", "<...
llama_model_loader: - kv  13:                      tokenizer.ggml.scores arr[f32,32000]   = [0.000000, 0.000000, 0.000000, 0.0000...
llama_model_loader: - kv  14:                  tokenizer.ggml.token_type arr[i32,32000]   = [2, 3, 3, 6, 6, 6, 6, 6, 6, 6, 6, 6, ...
llama_model_loader: - kv  15:                      tokenizer.ggml.merges arr[str,61249]   = ["▁ t", "e r", "i n", "▁ a", "e n...
llama_model_loader: - kv  16:                tokenizer.ggml.bos_token_id u32              = 1
llama_model_loader: - kv  17:                tokenizer.ggml.eos_token_id u32              = 2
llama_model_loader: - kv  18:            tokenizer.ggml.unknown_token_id u32              = 0
llama_model_loader: - kv  19:               tokenizer.ggml.add_bos_token bool             = true
llama_model_loader: - kv  20:               tokenizer.ggml.add_eos_token bool             = false
llama_model_loader: - kv  21:                    tokenizer.chat_template str              = {% if messages[0]['role'] == 'system'...
llama_model_loader: - kv  22:               general.quantization_version u32              = 2
llama_model_loader: - type  f32:   65 tensors
llama_model_loader: - type q4_0:  225 tensors
llama_model_loader: - type q6_K:    1 tensors
llm_load_vocab: special tokens definition check successful ( 259/32000 ).
llm_load_print_meta: format           = GGUF V3 (latest)
llm_load_print_meta: arch             = llama
llm_load_print_meta: vocab type       = SPM
llm_load_print_meta: n_vocab          = 32000
llm_load_print_meta: n_merges         = 0
llm_load_print_meta: n_ctx_train      = 4096
llm_load_print_meta: n_embd           = 4096
llm_load_print_meta: n_head           = 32
llm_load_print_meta: n_head_kv        = 32
llm_load_print_meta: n_layer          = 32
llm_load_print_meta: n_rot            = 128
llm_load_print_meta: n_embd_head_k    = 128
llm_load_print_meta: n_embd_head_v    = 128
llm_load_print_meta: n_gqa            = 1
llm_load_print_meta: n_embd_k_gqa     = 4096
llm_load_print_meta: n_embd_v_gqa     = 4096
llm_load_print_meta: f_norm_eps       = 0.0e+00
llm_load_print_meta: f_norm_rms_eps   = 1.0e-05
llm_load_print_meta: f_clamp_kqv      = 0.0e+00
llm_load_print_meta: f_max_alibi_bias = 0.0e+00
llm_load_print_meta: f_logit_scale    = 0.0e+00
llm_load_print_meta: n_ff             = 11008
llm_load_print_meta: n_expert         = 0
llm_load_print_meta: n_expert_used    = 0
llm_load_print_meta: causal attn      = 1
llm_load_print_meta: pooling type     = 0
llm_load_print_meta: rope type        = 0
llm_load_print_meta: rope scaling     = linear
llm_load_print_meta: freq_base_train  = 10000.0
llm_load_print_meta: freq_scale_train = 1
llm_load_print_meta: n_yarn_orig_ctx  = 4096
llm_load_print_meta: rope_finetuned   = unknown
llm_load_print_meta: ssm_d_conv       = 0
llm_load_print_meta: ssm_d_inner      = 0
llm_load_print_meta: ssm_d_state      = 0
llm_load_print_meta: ssm_dt_rank      = 0
llm_load_print_meta: model type       = 7B
llm_load_print_meta: model ftype      = Q4_0
llm_load_print_meta: model params     = 6.74 B
llm_load_print_meta: model size       = 3.56 GiB (4.54 BPW)
llm_load_print_meta: general.name     = LLaMA v2
llm_load_print_meta: BOS token        = 1 '<s>'
llm_load_print_meta: EOS token        = 2 '</s>'
llm_load_print_meta: UNK token        = 0 '<unk>'
llm_load_print_meta: LF token         = 13 '<0x0A>'
ggml_cuda_init: GGML_CUDA_FORCE_MMQ:   no
ggml_cuda_init: CUDA_USE_TENSOR_CORES: yes
ggml_cuda_init: found 1 ROCm devices:
  Device 0: AMD Radeon 780M Graphics, compute capability 11.0, VMM: no
llm_load_tensors: ggml ctx size =    0.22 MiB
llm_load_tensors: offloading 32 repeating layers to GPU
llm_load_tensors: offloading non-repeating layers to GPU
llm_load_tensors: offloaded 33/33 layers to GPU
llm_load_tensors:      ROCm0 buffer size =  3577.56 MiB
llm_load_tensors:        CPU buffer size =    70.31 MiB
..................................................................................................
llama_new_context_with_model: n_ctx      = 2048
llama_new_context_with_model: n_batch    = 512
llama_new_context_with_model: n_ubatch   = 512
llama_new_context_with_model: freq_base  = 10000.0
llama_new_context_with_model: freq_scale = 1
llama_kv_cache_init:      ROCm0 KV buffer size =  1024.00 MiB
llama_new_context_with_model: KV self size  = 1024.00 MiB, K (f16):  512.00 MiB, V (f16):  512.00 MiB
llama_new_context_with_model:  ROCm_Host  output buffer size =     0.14 MiB
llama_new_context_with_model:      ROCm0 compute buffer size =   164.00 MiB
llama_new_context_with_model:  ROCm_Host compute buffer size =    12.01 MiB
llama_new_context_with_model: graph nodes  = 1030
llama_new_context_with_model: graph splits = 2
ggml_cuda_compute_forward: RMS_NORM failed
CUDA error: invalid device function
  current device: 0, in function ggml_cuda_compute_forward at C:/a/ollama/ollama/llm/llama.cpp/ggml-cuda.cu:2212
  err
GGML_ASSERT: C:/a/ollama/ollama/llm/llama.cpp/ggml-cuda.cu:60: !"CUDA error"
time=2024-04-22T20:19:19.352+10:00 level=ERROR source=routes.go:120 msg="error loading llama server" error="llama runner process no longer running: 3221226505 CUDA error: invalid device function\r\n  current device: 0, in function ggml_cuda_compute_forward at C:/a/ollama/ollama/llm/llama.cpp/ggml-cuda.cu:2212\r\n  err\r\nGGML_ASSERT: C:/a/ollama/ollama/llm/llama.cpp/ggml-cuda.cu:60: !\"CUDA error\""
[GIN] 2024/04/22 - 20:19:19 | 500 |    5.1720784s |       127.0.0.1 | POST     "/api/generate"
4thanks commented 4 months ago

我按这个测试了ollama_windows_10_rx6600xt_zluda，amd显卡能成功运行并加速推理，你也试试。

kenhuang commented 4 months ago

Wow, confirm latest release works, did follow instruction copy cublas.dll to Ollama folder and renamed, 50% performance bump via APU IMG_2442 IMG_2443

kenhuang commented 4 months ago

Try use 5700xt getting errors, maybe related to my ROCm hack, it’s also not on support list. Happy with the APU working 🎉 IMG_2445 IMG_2444 IMG_2446

4thanks commented 4 months ago

Try use 5700xt getting errors, maybe related to my ROCm hack, it’s also not on support list. Happy with the APU working 🎉 !

我也是5700xt，zluda有支持并调用gpu，可以试试ROCmLibs_Testing.7z那个包，我用的没问题。